Journal of Controlled Release 65 (2000) 173–186

www.elsevier.com / locate / jconrel

Making microencapsulation work: conformal coating,

immobilization gels and in vivo performance q
a, a a b
M.V. Sefton *, M.H. May , S. Lahooti , J.E. Babensee
a
Institute for Biomaterials and Biomedical Engineering and Department of Chemical Engineering and Applied Chemistry,
University of Toronto, Toronto, Ontario M5 S 3 E5, Canada
b
Cox Laboratory, Institute of Biosciences and Bioengineering, Rice University, 6100 South Main, Houston, TX 77005, USA

Received 17 March 1999; accepted 29 August 1999

Abstract

Microencapsulation of cells as a means of insulin or other protein delivery (for example, for gene therapy) has not yet
realized its potential. Three aspects of this problem are illustrated with reference to the use of poly(hydroxyethyl
methacrylate-co-methyl methacrylate) (HEMA–MMA). Conformal coating was used to coat cell aggregates with a very thin
layer of a water-insoluble HEMA–MMA membrane that conforms to the shape of the aggregate, and minimizes the
polymer’s contribution to the total transplant volume. Cell aggregates were coated at a liquid–liquid interface of a
discontinuous density gradient composed of both aqueous and organic liquids. Aggregates of HepG2 cells were coated and
remained viable. Immobilization matrices were co-encapsulated in order to control cell phenotype. Ultralow gelling
temperature agarose promoted the proliferation of HEK293 cells, while the viability of transfected C2C12 cells was
improved in microcapsules that contained Matrigel  . Rat or human hepatoma cells in HEMA–MMA microcapsules lost
viability within a week after implantation into an omental pouch in Wistar rats. The loss of viability was attributed to the
tissue reaction, although it is not clear if the cells lost their viability in vivo leading to the aggressive tissue reaction or if the
latter caused the cells to starve or otherwise die. On the other hand, intraperitoneal implantation of microcapsules containing
L929 cells in ‘syngeneic’ C3H mice in a high-strength agarose gel resulted in maintenance of viability of |50% of the
encapsulated cells. While progress is being made on several fronts, this type of tissue engineering construct is still several
years away from routine use in humans.  2000 Elsevier Science B.V. All rights reserved.

Keywords: Microencapsulation; Conformal coating; Immobilization gels; Agarose

1. Introduction

But for one clinical report [1], microencapsulation

q
Presented at the Ninth International Symposium on Recent
Advances in Drug Delivery Systems, Salt Lake City, UT, USA,
22–25 February 1999.
*Corresponding author. Tel.: 11-416-978-3088; fax: 11-416-
978-4317.
E-mail address: sefton@ecf.utoronto.ca (M.V. Sefton)
of islets has not been implemented clinically, despite
almost 20 years since the first report [2]. While
reliable and reproducible isolation of pig islets
continues to be a technical challenge, microencapsu-
lation of cells has not been as ‘simple’ as originally
anticipated. The difficulties that have been overcome

0168-3659 / 00 / $ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S0168-3659( 99 )00234-5
174 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
are similar to those that are to be seen in other tissue
engineering constructs, of which microencapsulation
is but one example.
Microencapsulation in a polymer membrane is a
means of isolating cells from the immune system,
thereby enabling the transplantation of mammalian
cells without immunosuppression and the use of
xenogeneic or genetically engineered cells. The cells
are transplanted to correct a disease state by the
delivery of a cell product, typically a protein. Insulin
from pancreatic islets is an example of this mode of
therapy. Unlike conventional drug delivery devices,
the cells have an inexhaustible supply of the protein
(pending cell viability) in an intrinsically stable form
and without the problem and expense of protein
purification. Furthermore, the protein is delivered at
a rate determined by the normal physiology of the
cells which might involve regulation by glucose level
(islets) or potassium concentration (dopamine secret-
ing cells) or cytokine levels (antitrypsin).
Because microencapsulated cells are physically
isolated from the host, a degree of flexibility is
attained in the choice of cells that may be used for
implantation. For example, standard laboratory cell
lines that have been genetically engineered to secrete
a specific protein may be used to ameliorate a
disease state [3,4]. This approach offers an alter-
native to autologous somatic gene therapy where the
patient’s own cells are harvested, modified ex vivo,
and subsequently implanted [5,6]. Cell lines offer a
larger source of implantation tissue and as a result
increase the potential number of patients that may be
treated, and also facilitate the fulfilment of regulatory
and safety standards. In several experiments reported
here we used transfected cells as part of a larger
program exploring microencapsulated cells as a
platform for gene therapy.
Like other tissue engineering constructs, cell en-
capsulation depends on using polymers to control the
behavior of cells both before and after implantation.
Successful microencapsulation requires (1) high cell
number and viability, (2) control of cell function
(through the extracellular matrix, for example) and
(3) maintenance of function for extended duration
(Table 1). An adequate cell mass must be trans-
planted in a suitably small, appropriately shaped
volume. Diffusion limitations must be minimized to
ensure adequate nutrient (glucose / oxygen) supply,
Table 1
Requirements of a successful tissue engineering construct such as
microencapsulated cells
To be successful, the tissue To meet the need,
engineering construct needs must control
Adequate cell mass Diffusion limitations
Nutrient supply
Appropriate cell function Extracellular matrix
(phenotype) Cytokines
Sufficient duration Biocompatibility
Inflammation / immune system
metabolite efflux and product delivery. The cell
behavior must be controlled through the use of
appropriate extracellular matrices and by understand-
ing the effect of in situ generated cytokines. Finally,
the cell behavior must be sustained for several weeks
or months and this requires control of the host
response: even if the encapsulating membrane pre-
vents IgG from coming in contact with the cells, the
antigens shed by the cells can influence (adversely)
the system performance. These features of encapsu-
lated cells are illustrated in this review of recent
work with the use of hydroxyethyl methacrylate–
methyl methacrylate copolymer (HEMA–MMA). All
topics are described in greater detail elsewhere:
conformal coating [7], immobilization gels [8] and in
vivo behavior [9,10]. The use of HEMA–MMA to
microencapsulate cells and the in vitro performance
of the microencapsulated cells is reviewed elsewhere
[11].
While the problems of cell encapsulation are
illustrated here with respect to HEMA–MMA only,
these difficulties are generally applicable to other
microencapsulation systems, such as the more com-
monly used alginate–polylysine, or other immuno-
isolation systems such as hollow-fibre-based mac-
roencapsulation systems. For example, the latter are
limited by the inability to scale up, except by using
inordinately long fibres to get the desired cell mass
into an implantable volume. Alginate–polylysine
capsules appear to fail in vivo because of instability
after a few months in vivo, presumably due to loss of
calcium. The question of antigen shedding and
indirect antigen presentation (see Discussion) to
ultimately limit in vivo viability, without immune
modulation, is a general problem that applies to all
immunoisolation systems. However, the significance
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186 175

may vary depending on the cell type, cell source or
implant site.
1.1. Conformal coating
Conformal coating is a means of making very
small microcapsules that maximize the amount of
cells that can be delivered while minimizing the
device volume. In our hands, conformal coating
refers to a method that adds a 10–25 mm thick
coating of swollen polymer to a |150 mm cell
aggregate. Conformal coating also ensures that cells
are supplied with the maximum amount of nutrients.
Conformal coating has been implemented with rat
islets and human hepatoma cell aggregates [7].
Conformal coating involves the entrainment of poly-
mer solution around cell aggregates which occurs as
they approach (while being centrifuged) the interface
between the HEMA–MMA solution and the inter-
mediate layer (Fig. 1). Entrainment of HEMA–
MMA solution depended on the balance between two
processes: large deformation of the liquid–liquid
interface by the approaching aggregates, and slow
drainage of the thin film between the particles and
the interface. If the aggregate plus entrained polymer
solution reached the non-solvent layer [phosphate-
buffered saline (PBS) in our case] before the
HEMA–MMA solution drained from around the
particle, then a membrane precipitated around the
particle in the non-solvent layer. Interface deforma-
tion by small particles with specific gravity near
unity required that the interfacial tensions throughout
the apparatus [12], and especially at the coating
interface [13], were extremely low (,10 22 –10 23
dynes / cm), and that the gradient was centrifuged at
greater than 100 g. The former was achieved by
using the same solvent, polyethylene glycol 200
(PEG200), for all three organic layers; PEG200 is a
solvent for HEMA–MMA that can be tolerated by
cells, at least to a limited extent, despite its non-
aqueous nature [14].
1.2. Immobilization matrices
The physical isolation of microencapsulated cells
from the host allows modulation of the behavior of

Fig. 1. Schematic diagram of the conformal coating discontinuous density gradient. Representative specific gravities are shown for the upper
and lower layers, which are aqueous solutions of glycerol in PBS. The middle layer is a solution of 5–20% (w / w) HEMA–MMA in
PEG200. The other two intermediate layers are solutions of water-soluble polymers in PEG200.
176 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
the cells through alteration of their microenviron-
ment by inclusion of appropriate extracellular ma-
trices such as collagen or Matrigel in the core of the
capsule [15]. This matrix may provide physical
support and uniform distribution of the immobilized
cells such that local transport gradients of nutrients is
improved, buildup of metabolic waste products is
reduced, and necrosis is prevented. In addition, the
matrix may interact biochemically with receptors on
the surface of the encapsulated cells to induce
phenotypical changes in the cells. The effect of
inclusion of a matrix in the core of the capsules on
the behavior of two different transfected cell lines,
namely human embryonal kidney (HEK293) and
mouse C2C12 myoblasts, was investigated in vitro.
1.3. In vivo performance
The success of cell transplantation is determined
by the prolonged maintenance of cell viability and
expression of differentiated functions within the
implanted capsule. This functional success is depen-
dent, in part, on the minimization of tissue reaction
and fibrous capsule formation upon implantation.
Various mammalian cells have been implanted in
various sites in rats: subcutaneously, intramuscularly
(i.m.), free-floating within the peritoneal cavity,
within the omentum or an omental / i.m. pouch and in
the striatum. The nature of the tissue reaction
depended on the implant site, duration of implanta-
tion, the presence or absence of xenogeneic proteins
or other components (from the tissue culture serum,
the extracellular matrix or the cells themselves) and
the mode of implantation — as free-floating capsules
suspended in PBS or as capsules embedded in an
agarose disk. Here we summarize two studies, one
without agarose in an omental pouch and one with
intraperitoneal implantation of agarose disks.
2. Methods and materials
2.1. Cells
All cells were routinely maintained in antibiotic
(penicillin / streptomycin) and serum supplemented
medium as specified by ATCC for each cell type.
The mouse myoblasts were transfected with calcium
phosphate precipitated pNMG encoding the human
growth hormone (hGH, M.W. 27,100) gene (from
Prof. P.L. Chang, Department of Pediatrics, McMas-
ter University, Hamilton, Ontario, Canada). The
HEK293 cells were transfected to produce human
hepatic lipase with the pRc / CMV plasmid (Invit-
rogen, San Diego, CA, USA) by the calcium phos-
phate precipitation protocol (from the Lipid Research
Laboratory, St. Michael’s Hospital, Toronto, Ontario,
Canada). The L929 cells were transfected with
secretable alkaline phosphatase (SEAP) using the
MSEAPINV retrovirus (from Dr. Robert Hawley,
Toronto Hospital).
2.2. Cell encapsulation
All cells were encapsulated at a concentration of
5310 6 cells / ml. HEK cells were suspended in
medium that contained either 15% (w / v) Ficoll-400
(Pharmacia, Uppsala, Sweden) or 1% (v / v) ultralow
gelling temperature SeaPrep  (FMC) agarose (2%
w / v agarose solution in phosphate-buffered saline
was mixed with complete tissue culture medium at a
ratio of 1:1). The myoblasts were suspended in
medium that contained 15% (w / v) Ficoll-400 or 1–2
mg / ml Vitrogen 100, or 50% (v / v) Matrigel. The
L929 cells were suspended in 15% Ficoll, only.
Except for conformal coating (see below), encapsu-
lated hepatoma cells were intended for in vivo
studies. Thus to minimize serum transfer with the
cells, after trypsinization (and before encapsulation),
the washed cell pellet was suspended in Hormonally
Defined Medium [HDM: PC-1E Low Protein
Serum-Free Liquid base medium containing PC-1E
Sterile Supplement (Hycor Biomedical Inc., Irvine,
CA, USA), 2 mM L-glutamine (Gibco), 100 U / ml
penicillin and 100 ng / ml streptomycin (Gibco)]
before being suspended in 10% w / v Ficoll 400
(Sigma, St. Louis, MO, USA) in HDM. A similar
protocol was used with one series of experiments
with L929 cells, but with protein-free medium
(Sigma).
HEMA–MMA microcapsules, |400 mm in diam-
eter (e.g., 370617 mm by SEM), were prepared by
the submerged nozzle–liquid jet extrusion process as
described elsewhere [16,29]. Temperature control
measures were required for inclusion of collagen or
Matrigel in the capsule core. In order to prevent
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
gelation of the core solution prior to droplet forma-
tion, it was necessary to cool the syringe delivering
the cell suspension to the needle assembly. These
measures were not needed for agarose. After wash-
ing, capsules were maintained in HDM at 378C in a
humidified atmosphere of 5% CO 2 / 95% air for 8
days before implantation. The cell counts were
performed with samples of |40 microcapsules; cap-
sules were broken, the cells separated from the
debris and the cells counted by hemocytometer.
2.3. Conformal coating
HepG2 cells were cultured in suspension in 10 cm
polystyrene petri dishes for 3–5 days and sponta-
neously formed 50–400 mm spheroids. Liquids were
layered carefully in a 1.0 Ml tuberculin syringe
(Becton Dickinson, NJ, USA) with a plunger, but
with the stem removed. With cell aggregates in the
top layer, the gradient was centrifuged at 100–500 g
for 5 to 10 min. Cell aggregates were removed from
the gradient with a pasteur pipette into PBS or cell
medium for further curing and subsequent analysis.
The syringe was capped and then cut with centrifuge
tube cutters to expose the bottom layer so that the
coated cell aggregates could be removed. For cell
aggregate coating, the gradient consisted of (from
bottom to top): 0.3 ml of an aqueous solution
consisting of 60% (w / w) glycerol, 0.05 ml 35%
(w / w) polyvinyl pyrrolidone (M.W. 2.5 kD), 0.05 ml
10% (w / w) HEMA–MMA in PEG200, 0.05 ml pure
PEG200, 0.2 ml a-MEM1antibiotics110% fetal
bovine serum and 200–300 HepG2 cell aggregates.
After coating, cells were fixed as below, gold coated
and examined by SEM.
2.4. Microcapsule implantation and recovery
For omental pouch implants, an aliquot of 200
hepatoma cell capsules (8 days post-encapsulation)
were washed with PBS and implanted into Wistar
rats (175–200 g, Charles River) in omental pouches
(200 capsules / pouch / rat) for 1, 4, 7, and 14 days.
The capsules were recovered within the omental
tissue into which they had been placed, washed with
PBS and then fixed [3.5% (v / v) glutaraldehyde, EM
grade (Polysciences, Warrington, PA, USA), con-
taining 2% (w / v) tannic acid in 0.1 M Sorenson’s
177
phosphate buffer (P-buffer)]. Non-encapsulated cells
were not tested, since the absence of a capsule
precludes finding the cells for histological evalua-
tion. Further details of the implantation and recovery
protocols are presented elsewhere [9].
Microcapsules containing transfected L929-SEAP
cells were also implanted 8 days post-encapsulation
intraperitoneally in male 5–6 week old ‘syngeneic’
C3H mice (20–24 g, Charles River) through a 1–2
cm abdominal mid-line incision. The capsules were
separated into aliquots of 200 capsules, washed and
then embedded in a 3 mm thick agarose disk
consisting of either 5% (w / v) SeaPrep  (ultralow
gelling, 400 g / cm 2 gel strength, 15 mm diameter
disk) or 4% (w / v) SeaPlaque  (low gelling, 1250
g / cm 2 gel strength, 7 mm diameter disk); both
products from FMC. Explantations were typically
performed on days 3, 10, and 21 post-implantation.
The agarose disk was transferred to fresh PBS and
the capsules were released from the gel by cutting
the gel apart with a surgical blade. The released
microcapsules were separated from the sliced aga-
rose gel pieces and host tissue with a 70 mm nylon
cell strainer and were transferred to the appropriate
medium.
2.5. Microscopy
Tissue samples containing capsules and capsules
maintained in vitro were prepared for light micro-
scopy as previously described [9,28]. Serial cryostat
sections (5–8 mm in thickness) were stained with
either aqueous toluidine blue [0.1% (w / v) in distilled
water, BDH Chemicals, Toronto, Ontario, Canada],
Masson trichrome (Sigma Chemical Company) or
Harris haemotoxylin and alcoholic eosin (Sigma
Chemical Company). Morphometric analysis of the
tissue reaction to capsules and the encapsulated cell
morphology was performed with Image 1 image
analysis software (Version 4.0, Universal Imaging
Corporation, West Chester, PA, USA) connected to a
Zeiss Axiovert microscope via a 3CCD Video Ca-
mera (Model DXC-930, Sony). The quantification
scheme is described elsewhere [9].
Confocal scanning laser images were obtained by
staining conformally coated cells with 0.05 mg / ml
ethidium homodimer and 0.25 mg / ml of calcein AM
(both from Molecular Probes, Eugene, OR, USA) for
178 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186

30 min in cell medium and then rinsing twice with
PBS. Images represent an optical cross-section
through the center of the coated cell aggregates and
were obtained using an MRC-600 confocal scanning
laser microscope.
3. Results and discussion
3.1. Conformal coating
Fig. 2 shows two spherical HepG2 cell aggregates
conformally coated in a single HEMA–MMA mem-
brane. Individual cells of the aggregates are visible
through a defect in the coating. Fig. 3 shows the
presence of a thin HEMA–MMA coating around
viable HepG2 cell aggregates shortly after passage
through the density gradient and exposure to the
organic components.
The gradient was modified to coat cells because
their specific gravity increased (to .1.18 g / ml) from
exposure to PEG200. Consequently, cell aggregates
were coated as they passed from the top of the
gradient to the bottom, and the coating interface

Fig. 2. Scanning electron microscope image of a thin conformal coating around HepG2 cell aggregates.
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186 179

Fig. 3. Confocal scanning laser microscope images of conformally coated HepG2 cell aggregate indicating that the cell aggregates are viable
after being coated. Original colored image was digitized and converted to a grey-scale to permit publication.

became the interface between HEMA–MMA solu-
tion and the lower intermediate layer. The increase in
cell aggregate specific gravity indicated that the
PEG200 solvent dehydrated the cell aggregates. The
long-term effects on viability and function of this
dehydration, as well as the rehydration that occurs in
the precipitation layer, are being investigated. HepG2
spheroids cultured for coating were relatively dense
and tightly packed. The coating process, however,
did not appear limited by minor differences in size,
shape, or packing of the HepG2 cell aggregates; the
process depended fundamentally on the presence of a
‘particle’ to deform the coating interface.
Conformal coating at a liquid–liquid interface
shows promise as a method to produce ultra-thin
conformal coatings of synthetic polymers around
small particles and cell aggregates. Whether the
membrane has the necessary permselectivity to iso-
late cells from the immune system or whether the
coated cells maintain their functional attributes in
vivo are under investigation.
3.2. Effect of immobilization /attachment matrices
Cell growth in the presence and absence of a
matrix is shown in Fig. 4 for both HEK and C2C12
cells. Agarose immobilization was beneficial for
transfected HEK cells, while collagen or Matrigel
were unable to preserve completely the viability of
encapsulated transfected C2C12 cells. Matrigel was
better than collagen, presumably, in part, because the
latter did not undergo contraction. In the absence of
agarose, there was little change in cell number (per
capsule) for untransfected HEK cells, while the cell
numbers for transfected ones decreased steadily. The
number of live 293pHL9 cells decreased from |100
cells / capsule on day 1 to #20 cells / capsule 10 days
post-encapsulation. Transfection had altered cell
behavior so that viability could not even be main-
tained. Within a day after encapsulation without
agarose, the free-floating individual cells had formed
a single aggregate within the core of the capsule. It
appears that because poly(HEMA–MMA) does not
180 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
Fig. 4. Effect of co-encapsulated immobilization matrices on
encapsulated cell number in vitro. (a) HEK293 cells with (h) or
without (d, j) 1% SeaPrep agarose. (b) C2C12 cells without a
matrix (d) or with 1.5 mg / ml collagen (n) or 50% Matrigel (s).
293, untransfected HEK cells; 293pHL9, transfected HEK293
cells; C2C12-hGH, transfected C2C12 cells.
support the adhesion of cells, the anchorage-depen-
dent HEK cells established cell–cell contacts in
order to maintain viability. However, this was in-
sufficient to promote the proliferation of the untrans-
fected HEK cells or, in the case of the transfected
cells, to prevent cell death. In the presence of 1%
(w / v) ultralow gelling temperature SeaPrep agarose,
however, the transfected HEK cells gave rise to a
multitude of aggregates within the gel through
proliferation. The initial |200 cells / capsule doubled
within 14 days after encapsulation and reached a
plateau value below 500 cells / capsule at the end of
the 28 day observation period.
The lack of cell adhesion to HEMA surfaces has
been well established. The formation of aggregates
by anchorage-dependent cells in a non-adhesive
environment has also been well documented and
described for tumour spheroids [17,18]. As tumour
spheroids grow in size, the presence of gradients of
critical nutrients and metabolic wastes cause de-
velopment of central necrosis while the cells on the
periphery consist of both proliferating and quiescent
cells. Within an aggregate, the core cells suffer from
a more limited access to the supply of nutrients
compared to the cells on the periphery. Thus, the
formation of a single aggregate in the capsule core
limited the number of cells that were able to thrive
within the poly(HEMA–MMA) microenvironment.
As a result, all the cell types (except the untransfect-
ed HEK) suffered a drastic loss in viability with the
passage of time and were not able to thrive based on
the cell–cell contacts that were established within the
aggregate.
Co-encapsulation with agarose not only main-
tained the viability of the encapsulated 293pHL9
cells, but also allowed for their proliferation. Im-
mobilization of the cells in agarose provided a
uniform distribution of the cells throughout the
capsule core. As a result, the diffusion of nutrients to
the cells was improved and remained unhindered by
the presence of any neighboring cells. Scanning
electron micrographs (not shown) suggested that the
proliferating aggregates existed within a milieu of an
extracellular matrix. Thus, the agarose gel allowed
the encapsulated cells to modulate their microen-
vironment by acting as a substrate for deposition of
extracellular proteins.
In the absence of a matrix, the encapsulated
C2C12-hGH cells (and untransfected cells) also
formed a single aggregate in the core of the capsule.
Here, however, the size of the aggregate (and the cell
count per capsule) decreased, indicating a progres-
sive decline in the number of viable cells. Although
collagen was a good substrate for the cells in
conventional tissue culture, only |20% of the cells
remained viable after 14 days. On the other hand,
co-encapsulation with 50% (v / v) Matrigel at a cell
density of 5310 6 cells / ml improved the percentage
of viable C2C12-hGH cells to |50% 21 days post-
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
encapsulation. While the collagen gels underwent
significant contraction, the 50% (v / v) Matrigel ma-
trix did not.
It was expected that co-encapsulation of myoblasts
with a collagen matrix, which supported the cultiva-
tion of the cells in culture, would improve the
viability of the encapsulated myoblasts by providing
specific sites of adhesion for the cells. However, the
collagen capsules were unsuccessful. The collagen
gel in the core of the capsules underwent contraction,
and the majority of the live encapsulated cells
existed on the outer surface of the gel rather than
within the core of the gel. The compaction of the
encapsulated cells in a smaller volume within the
core decreases the local availability of nutrients and
increases the competition for the available nutrient
supply. Thus, the improvement in local transport
kinetics of nutrients that is gained from a uniform
distribution of the cells throughout the core of the
capsule by co-encapsulation with a matrix is negated.
Evidence from the preparation of hybrid muscular
tissues, where C2C12 cells were incorporated within
collagen matrices, indicates that the contraction of
the collagen matrix and geometrical considerations
that dictate the transport kinetics of nutrients com-
bine to determine the viability of the cells within the
matrix [19].
On the other hand, the Matrigel matrices were not
contracted by the myoblasts. The preferential growth
(or rather reduced loss of viability) of the C2C12
cells in Matrigel may also be explained by the
presence of other extracellular matrix proteins in
addition to collagen [20]. Unfortunately, the presence
of a mixture of mouse proteins limits the applicabili-
ty of this matrix for in vivo implantations.
There are several examples of inclusion of ex-
tracellular or immobilization matrices in encapsula-
tion devices. These include co-encapsulation of
transfected baby hamster kidney cells (BHK) secret-
ing human nerve growth factor (hNGF) with col-
lagen in poly(acrylonitrile–vinyl chloride) (PAN–
PVC) hollow fibre macrocapsules [21] and transfect-
ed mouse Ltk – fibroblasts secreting human growth
hormone (hGH) with collagen in alginate–poly-L-
lysine–alginate spherical microcapsule [22]. The use
of Matrigel in HEMA–MMA microcapsules is dis-
cussed in detail elsewhere [23,24]. Rat islets have
been encapsulated with agarose in AN69 [poly(acryl-
181
onitrile–sodium methallylsulfonate)] fibres [25] and
rat pheochromocytoma PC-12 cells have been used
with precipitated chitosan in PAN–PVC fibres
[26,27]. Although agarose does not provide specific
sites for the adhesion of cells, it distributes the cells
more uniformly within the capsule core to reduce
nutrient diffusion limitations.
3.3. Microencapsulated hepatoma cells in vivo —
omental pouch
Implanted into an omental pouch without an
agarose disk, microencapsulated rat hepatoma cells
(H4IIEC3) remained viable longer than did human
hepatoma cells (hepG2), although microencapsulated
cell viability was limited in both cases. Based on
morphological characteristics, 25–50% of the mi-
croencapsulated rat hepatoma cells were viable at 7
days, while 25–50% viability was seen only up to
day 4 for the microencapsulated human hepatoma
cells. As early as 4 days after implantation, both
types of microencapsulated cells showed signs of cell
degeneration.
Human hepatoma cell viability and morphology in
microcapsules, 4 days in vivo, appeared to correlate
with the degree of vascularization of the tissue
reaction in omental tissue. Microcapsules, within
omental tissue, with a well vascularized tissue re-
action (Fig. 5a), contained viable human hepatoma
cells (Fig. 5b), while those with poorly vascularized
tissue reactions did not (Fig. 5c). This accelerated
loss of viability (relative to what was seen in vitro) is
likely the consequence of the tissue reaction: the
additional diffusion limitations, the consumption of
nutrients by the granulation tissue or its production
of cytokines, reactive oxygen intermediates or pro-
teases. The critical feature is not known. The detailed
characteristics of the host reaction are reported
elsewhere [9].
Despite the overall similarities, there were several
distinguishing characteristics of the tissue reactions
to microcapsules containing the two types of cells.
With human hepatoma cells tissue reaction thick-
nesses shifted to higher values than with the rat
hepatoma cells. For example, at day 7 the median
thickness was 119 mm for HepG2 cell implants and
55 mm for the rat cells. Furthermore, there appeared
to be more eosinophils associated with microcapsules
182 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
Fig. 5. HEMA–MMA microcapsules, within omental tissue in
Wistar rats, with many vascular structures in the tissue reactions
(a) at day 4, contained viable HepG2 cells, while those with
poorly vascularized tissue reactions contained necrotic cells.
Magnification: bar, 50 mm. Staining: Masson trichrome (a) and
(b); hemotoxylin and eosin (c).
containing human cells as compared to microcap-
sules containing rat cells. The human HepG2 cells
were used here as the model cell for xenogeneic
transplantation. They have been used to show that
HepG2 cells survived entrapment within large [28]
and small diameter [30] HEMA–MMA capsules, and
were viable for up to 3 weeks when coencapsulated
with the cell attachment substrate Matrigel. Here,
they were implanted without Matrigel. H4IIEC3 cells
are a well differentiated rat hepatoma cell line [31–
33] derived from ACI rats [34,35]; they were used as
a model for allograft transplantation. Because of
differences in the rejection mechanism of allogeneic
and xenogeneic cells, it was expected that donor
differences would be apparent in the tissue reaction
to these microcapsules; such differences are detailed
elsewhere [9].
The tissue reaction that was seen was substantially
greater than that expected based on implants of
polymer alone (i.e., without cells) [36]. Care was
taken here to avoid the coencapsulated extracellular
matrix or the incubation with bovine serum proteins
which were thought to be important mediators of an
enhanced tissue response that was seen when these
(e.g., Matrigel) were used [25]. Nevertheless, avoid-
ing these xenogeneic proteins was not sufficient to
render the tissue reaction similar to that seen by the
polymer alone. We presume that the encapsulated
cells contributed to the tissue response through an
antigen shedding mechanism.
The original premise of immunoisolation was to
physically separate allogeneic or xenogeneic tissue
from the host immune system by preventing contact
with the host immune system. However, capsules
such as HEMA–MMA, which have a molecular
weight cut-off sufficient to exclude antibodies, do
not appear to prevent an immune response towards
antigens shed by the encapsulated cells. These
antigens may be foreign proteins secreted by the
cells including the therapeutic agent, cell surface
molecules (e.g., MHC molecules) shed as part of
normal behavior or cell components released upon
cell death. An immunological reaction against these
antigens may be produced via indirect antigen pre-
sentation (by host cells) [37] as the antigens per-
meate through the polymer capsule wall. Antigen
shedding from microcapsules would sensitize the
host to the transplanted cells and initiate a cytotoxic
tissue response to the implant, which would directly
compromise encapsulated cell viability and function.
Such a reaction would be expected to be stronger the
greater the degree of discordance between the donor
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186 183

and host, so that xenogeneic tissue would be more
likely to be destroyed than allogeneic tissue.
On the other hand, the tissue reaction may follow
the loss of viability. In vitro, the nutrient levels and
pO 2 are high, while in vivo the capsules are packed
relatively closely together. The combination of lower
concentrations and effectively higher cell densities
exacerbates the competition for nutrients. Then the
progressive loss of viability, due to starvation, may
result in the initiation of a stronger immune / in-
flammatory response. Dispersing capsules in agarose
may be a means of circumventing this problem.
3.4. Intraperitoneal implantation — use of agarose
disks
While the omental pouch facilitated the recovery
of the capsules, the close proximity of the capsules
to each other led to capsule agglomeration. This in
turn exacerbated the diffusion limitations that ulti-
mately led to premature cell death. In a subsequent
study, microencapsulated L929-SEAP cells (more
robust than the hepatoma cells, especially in the
absence of an extracellular matrix) were implanted
intraperitoneally in C3H mice. Since the C3H mice
are syngeneic for L929 cells, the main reason for the
death of the encapsulated cells is the diffusion
limitations that arise upon implantation. The capsules
were implanted as a suspension in PBS, but the
recovery was poor and most of the capsules were
found deformed in shape and in larger agglomerates.
Cell viability was low. Implantations were then
performed by using agarose disks. The cell counts
per capsule at different explantation times are shown
in Fig. 6 for both the low-strength, SeaPrep, and
higher-strength, SeaPlaque, disks. For the former
disks, the capsules were incubated in protein-free
medium prior to implantation to reduce the potential
sources of an immune response against the implant.
Further details are provided elsewhere [38].
One day prior to the implantations, i.e. day 7
post-encapsulation, there were 484633 live cells /
capsule. Within 3 days of implantation, the recovered
cell number had decreased to 378662 with the
lower-strength agarose and by 21 days there were
virtually ,30 viable cells per capsule (in the few
capsules that could be recovered). Within 3 days, the

Fig. 6. Comparison of the effect of lower-strength SeaPrep agarose (5%) and higher-strength SeaPlaque agarose (4% ck percentages) on the
preservation of encapsulated L929-SEAP cell viability for 3 weeks of implantation in C3H mice. For the SeaPrep study, capsules were
maintained in protein-free and serum-free medium. For SeaPlaque, capsules were maintained in medium with serum.
184 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
lower-strength agarose gel (5% w / v SeaPrep) had
physically disintegrated under the internal stresses of
the peritoneal cavity into pieces that were 100–200
mm in size. By 21 days the agarose had completely
disappeared, and very few capsules were recovered.
Those that were, were in the form of aggregates
containing 15–90 microcapsules. Presumably be-
cause the capsules were exposed directly to the
stresses inside the peritoneal cavity, within 3 days
the capsules lost their spherical geometry and had
become flattened, but there were no signs of break-
age or tearing of the poly(HEMA–MMA) mem-
brane. The released microcapsules were surrounded
by a single layer of host cells. The aggregation of the
microcapsules and the presence of the fibrous tissue
presumably posed limitations on the diffusion of
nutrients to the encapsulated cells, especially within
microcapsules that were within the core of the
aggregate.
While the low-strength SeaPrep agarose was un-
successful in solving the deformation, aggregation,
and retrieval problems that were encountered, the
higher-strength SeaPlaque agarose proved to be more
useful. The higher-strength agarose gel was free of
any attachments to internal organs, remained stable
and intact at all explantation time points. The
physical stability of the gel in vivo resulted in a
complete recovery of all the implanted microcap-
sules. In addition, a significant majority of the
microcapsules maintained their spherical shape and
were free of any deformations. By prevention of
their aggregation and deformation, it was possible to
maintain viability of the implanted cells for at least
21 days. Nevertheless, a reduction in the number of
viable cells / capsule was detected with the passage of
time (Fig. 6), although the reduction was less than
for the low-strength agarose. Once the viable cells
were released from retrieved microcapsules and
regrown as monolayers, they expressed SEAP at a
similar level to their encapsulated but non-implanted
counterparts [38]. At explantation, the agarose was
surrounded by a single layer of host cells. The host
cells did not penetrate the agarose gel and their
presence was restricted to the outer surface of the
disk. These microcapsules had not been washed or
maintained in protein-free medium prior to implanta-
tion. Hence, a more severe fibrous tissue reaction
occurred on the side of some gels where microcap-
sules were not completely enclosed since there was
direct exposure of the microcapsule membranes with
their adsorbed serum proteins to the peritoneal fluid.
One disadvantage of immobilization of the mi-
crocapsules in agarose is the introduction of a barrier
in addition to the poly(HEMA–MMA) membrane for
diffusion of nutrients from the peritoneal fluid to the
encapsulated cells. It would be interesting to implant
unencapsulated cells in agarose so as to distinguish
the effect of the agarose from that of the capsule. A
study of the viability and function of islets immobil-
ized in agarose gels showed that spherical implants
were superior to tubular, and disk-shaped implants
[39]. Optimization of the geometry and size of the
SeaPlaque agarose disk may lead to a further im-
provement of the viability of the encapsulated cells.
Similarly, strategies that increase the permeability of
the microcapsule membrane or the scaffold that is
used to deliver the microcapsules may improve the
viability of the encapsulated cells.
4. Conclusions
Microencapsulation of mammalian cells is poten-
tially a powerful means of delivering therapeutically
important molecules such as insulin. It can also have
numerous applications as a platform for gene
therapy. However, realizing this potential has been
more difficult than first anticipated. Small amounts
of protein per cell necessitates high cell numbers.
This, in turn, requires ultrasmall capsules such as
those produced by conformal coating if the total
device volume is to be kept small. Immobilization
matrices need to be added to the cells to minimize
intercellular diffusion gradients, while specific ex-
tracellular molecules may be needed to facilitate the
growth of anchorage-dependent cells. Finally, the
behavior of cells in vivo is critically dependent on
minimizing antigen shedding and on inhibiting the
corresponding presentation of these antigens in a
functional manner to host lymphocytes. This then
requires minimization of the host inflammatory
response and preservation of the viability of the
encapsulated cells as well as minimization of the use
of foreign proteins, for example in the serum or
matrices used with the cells. The agarose disk
technique may prove to be useful, if not in the actual
 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
device used in humans, at least in assessing the fate
of encapsulated cells in animal models. This method
may be especially useful for defining the benefit to
be gained by particular strategies designed to inter-
fere with cellular inflammatory or immune responses
to the capsules. Much work is needed if encapsulated
cell delivery is to become a useful means of con-
trolled release.
Immunoisolation in general and microencapsula-
tion in particular has proven to be more difficult than
originally anticipated. The alginate–polylysine sys-
tem is certainly simpler and the materials are more
readily available than HEMA–MMA. Hence it has
been developed further than the HEMA–MMA
system and has been subject to greater in vivo and
even limited clinical investigation. Nonetheless, nei-
ther system has had the expected impact. The
specific in vivo limitations are different — e.g.,
clumping for HEMA–MMA, loss of calcium for
alginate–polylysine — but the conclusion is the
same. Immunoisolation must await the development
of better materials and processes and better strategies
for controlling the immune response, if the promise
of this form of drug delivery is to be realized.
Acknowledgements
We acknowledge the financial support of the
Natural Sciences and Engineering Research Council
(NSERC) of Canada, the Medical Research Council
of Canada, the Ontario Centre for Materials Research
and the Liver Foundation of Canada. The latter
provided a fellowship to JEB, while NSERC pro-
vided graduate fellowships to MHM and SL. We are
also grateful for the technical assistance of Chuen Lo
and Susan Wang.
References
[1] P. Soon-Siong et al., Lancet 343 (1994) 950–951.
[2] F. Lim, A.M. Sun, Science 210 (1980) 908–910.
[3] N.A.-M. Pochon et al., Clinical protocol: gene therapy for
amyotrophic lateral sclerosis (ALS) using a polymer en-
capsulated xenogenic cell line engineered to secrete hCNTF,
Hum. Gene Ther. 7 (1996) 851–860.
[4] P.L. Chang, N. Shen, A.J. Westcott, Delivery of recombinant
185
gene products with microencapsulated cells in vivo, Hum.
Gene Ther. 4 (1993) 433–440.
[5] I.D. Dube, D. Cournoyer, Gene therapy: here to stay, Can.
Med. Assoc. J. 152 (1995) 1605–1613.
[6] A. Bank, Human somatic cell gene therapy, BioEssays 18
(1996) 999–1007.
[7] M.H. May, M.V. Sefton, Conformal coating of small particles
and cell aggregates at a liquid–liquid interface, Ann. New
York Acad. Sci. (in press).
[8] S. Lahooti, M.V. Sefton, Effect of an immobilization matrix
and capsule membrane permeability on the viability of
encapsulated HEK cells (in preparation).
[9] J.E. Babensee, M.V. Sefton, Transplantation of HEMA–
MMA microencapsulated hepatoma cells into rats, Cell
Transplant. (in press).
[10] S. Lahooti, M.V. Sefton, Characterization of microencapsu-
lated genetically engineered mouse fibroblasts (L929) im-
planted in the peritoneal cavity of C3H mice (in preparation).
[11] J.E. Babensee, M.V. Sefton, Protein delivery by microen-
capsulated cells, in: K. Park (Ed.), Protein Delivery: Chal-
lenges and Strategies, ACS Professional Reference Book,
ACS, Washington, DC, 1997, pp. 311–332.
[12] H.C. Maru, D.T. Wasan, R.C. Kintner, Behaviour of a rigid
sphere at a liquid–liquid interface, Chem. Eng. Sci. 26
(1971) 1615–1628.
[13] A.S. Geller, S.H. Lee, L.G. Leal, The creeping motion of a
spherical particle normal to a deformable interface, J. Fluid
Mech. 169 (1986) 27–69.
[14] C.A. Crooks, J.A. Douglas, R.L. Broughton, M.V. Sefton,
Microencapsulation of mammalian cells in a HEMA–MMA
copolymer: effects on capsule morphology and permeability,
J. Biomed. Res. 24 (1990) 1241.
[15] A.C. Jen, M.C. Wake, A.G. Mikos, Review: hydrogels for
cell immobilization, Biotechnol. Bioeng. 50 (1996) 357–
364.
[16] S. Lahooti, M.V. Sefton, Methods for microencapsulation
with HEMA–MMA, in: J.R. Morgan, M.L. Yarmush (Eds.),
Tissue Engineering Methods and Protocols, Humana Press,
Totowa, NJ, 1998, pp. 331–348.
[17] R.M. Sutherland, Cell and environment interactions in
tumour microregions: the multicell spheroid model, Science
240 (1988) 177–184.
[18] W.J. Mueller-Klieser, Multicellular spheroids: a review on
cellular aggregates in cancer research, Cancer Res. Clin.
Oncol. 113 (1987) 101–122.
[19] T. Okano, T. Matsuda, Tissue engineered skeletal muscle:
preparation of highly dense, highly oriented hybrid muscular
tissues, Cell Transplant. 7 (1998) 71–82.
[20] H.K. Kleinman et al., Isolation and characterization of type
IV procollagen, laminin, and heparan sulfate proteoglycans
from the EHS sarcoma, Biochemistry 21 (1982) 6188–6193.
[21] S.R. Winn et al., Polymer-encapsulated cells genetically
modified to secrete human nerve growth factor promote
survival of axotomized septal cholinergic neurons, Proc.
Natl. Acad. Sci. USA 91 (1994) 2324–2328.
[22] I.T. Tai, A.M. Sun, Microencapsulation of recombinant cells:
a new delivery system for gene therapy, FASEB J. 7 (1993)
1061–1069.
186 M.V. Sefton et al. / Journal of Controlled Release 65 (2000) 173 – 186
[23] J.E. Babensee, U. DeBoni, M.V. Sefton, Morphological
assessment of hepatoma cells (HepG2) microencapsulated in
a HEMA–MMA copolymer with and without Matrigel, J.
Biomed. Mater. Res. 26 (1992) 1401–1418.
[24] H. Uladag, M.V. Sefton, Microencapsulated human hepatoma
(HepG2) cells: in vitro growth and protein release, J.
Biomed. Mater. Res. 24 (1993) 1213–1224.
[25] P. Prevost et al., Application of AN69 hydrogel to islet
encapsulation: evaluation in the streptozotocin-induced dia-
betic rat model, Transplant. Proc. 27 (1995) 3393–3395.
[26] B.A. Zielinski, P. Abishcher, Chitosan as a matrix for
mammalian cell encapsulation, Biomaterials 15 (1994)
1049–1056.
[27] D.F. Emerich et al., Transplantation of polymer encapsulated
PC12 cells: use of chitosan as an immobilization matrix, Cell
Transplant. 2 (1993) 241–249.
[28] J.E. Babensee, U. De Boni, M.V. Sefton, Morphological
assessment of hepatoma cells (HepG2) microencapsulated in
a HEMA–MMA copolymer with and without Matrigel, J.
Biomed. Mater. Res. 26 (1992) 1401–1418.
[29] H. Uludag, M.V. Sefton, Microencapsulated human hepatoma
cells: in vitro growth and protein release, J. Biomed. Mater.
Res. 27 (1993) 1213–1224.
[30] H. Uludag, V. Horvath, J.P. Black, M.V. Sefton, Viability and
protein secretion from human hepatoma (HepG2) cells
encapsulated in 400 mm polyacrylate microcapsules by
submerged nozzle–liquid jet extrusion, Biotechnol. Bioeng.
44 (1994) 1199–1204.
[31] J.A. Scheider, M.C. Weiss, Expression of differentiated
functions in hepatoma cell hybrids, I. Tyrosine aminotrans-
ferase in hepatoma–fibroblast hybrids, Proc. Natl. Acad. Sci.
68 (1971) 127–131.
[32] S. Babajko, F. Tronche, A. Groyer, Liver-specific expression
of human insulin-like growth factor binding protein 1:
functional role of transcription factor HNF1 in vivo, Proc.
Natl. Acad. Sci. 90 (1993) 272–276.
[33] M. Takano et al., Interleukin-6 as a mediator responsible for
inflammation-induced increase in plasma angiotensinogen,
Biochem. Pharmacol. 45 (1993) 201–206.
[34] E.B. Thompson, G.M. Tomkins, J.F. Curran, Induction of
tyrosine a-ketoglutarate transaminase by steroid hormones in
a newly established tissue culture cell line, Biochemistry 56
(1966) 296–303.
[35] M.D. Reuber, A transplantable bile-secreting hepatocellular
carcinoma in the rat, J. Natl. Cancer Inst. 26 (1961) 891–
897.
[36] J.E. Babensee, U. De Boni, M.V. Sefton, Tissue reaction to
HEMA–MMA capsules: effect of device components (in
preparation).
[37] M.H. Sayegh, B. Watschinger, C.B. Carpenter, Mechanisms
of T cell recognition of alloantigen. The role of peptides,
Transplantation 57 (1994) 1295–1302.
[38] S. Lahooti, In vitro and in vivo characterization of ge-
netically engineered cells microencapsulated in a HEMA–
MMA copolymer, Ph.D. Thesis, University of Toronto, 1999.
[39] H. Yang et al., Comparative studies of in vitro and in vivo
function of three different shaped bioartificial pancreases
made of agarose hydrogel, Biomaterials 15 (1994) 113–120.